CN114492967A - Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model - Google Patents

Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model Download PDF

Info

Publication number
CN114492967A
CN114492967A CN202210047361.6A CN202210047361A CN114492967A CN 114492967 A CN114492967 A CN 114492967A CN 202210047361 A CN202210047361 A CN 202210047361A CN 114492967 A CN114492967 A CN 114492967A
Authority
CN
China
Prior art keywords
data
passenger flow
prediction
blstm
ceemdan
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210047361.6A
Other languages
Chinese (zh)
Inventor
王敏
翟佑春
周涛
董小彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hohai University HHU
Original Assignee
Hohai University HHU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hohai University HHU filed Critical Hohai University HHU
Priority to CN202210047361.6A priority Critical patent/CN114492967A/en
Publication of CN114492967A publication Critical patent/CN114492967A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Economics (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a method and a medium for predicting urban rail station passenger flow based on CEEMDAN and BLSTM combined models. The method comprises the steps of firstly adopting a CEEMDAN algorithm to decompose passenger flow data, then selecting BLSTM improved by LSTM to obtain the time sequence characteristics of a decomposition sequence based on the time sequence change of the passenger flow, and finally adding all component prediction results to obtain final passenger flow prediction data.

Description

Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model
Technical Field
The invention relates to the field of urban rail transit passenger flow prediction, in particular to an urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined models.
Background
Due to the characteristics of large urban rail transit carrying capacity, convenience, rapidness and the like, the role of urban traffic in playing is becoming more and more important. The method has the advantages that the accurate prediction of the change of passenger flow at the future moment has important significance for reasonably arranging operation plans and improving service levels of a track operation department, and meanwhile, the predicted passenger flow data can also provide support for the management department in pedestrian flow dispersion and safety precaution in a personnel-intensive environment. The air conditioner load in the subway station is influenced by the passenger flow in the station, the passenger flow data obtained by prediction and the relevant characteristics of temperature, humidity and the like are used as the basis for predicting the air conditioner load in the station, the air conditioner can be further subjected to energy-saving control after the air conditioner load prediction data is obtained, and the method has great significance for realizing the electric energy saving and the operation cost of the station.
The passenger flow prediction is to predict the passenger flow at the next moment or even a plurality of future moments by using historical passenger flow data and relevant influence factors, and the prediction time scale is generally 15 minutes, namely the passenger flow in the future 15 minutes is predicted. At present, passenger flow prediction methods based on statistical theory mainly have time sequence, linear regression models, Kalman filtering and the like. Although the conventional statistical theory method has been developed well, the method cannot meet the current requirement for the passenger flow prediction accuracy in view of the nonlinearity and the volatility of the passenger flow data.
With the progress of scientific technology and the development of big data, a prediction model based on machine learning is widely used, so that a passenger flow prediction method is more possible. The BP neural network is used for short-term traffic flow prediction by scholars as early as 1994, then long-short term memory neural network (LSTM) is used for passenger flow prediction by scholars and compared with other methods, the superiority of the LSTM is verified, the bidirectional long-short term memory neural network (BLSTM) is improved from the LSTM and is widely applied to time sequence data prediction, the bidirectional long-short term memory neural network is composed of two layers of LSTMs which are transmitted in the positive and negative directions, time sequence characteristics of the data in the positive and negative directions can be learned, and therefore higher prediction accuracy can be obtained.
Since passengers have certain randomness during traveling, the variation of the passenger flow volume of the station is also random and fluctuating, and a learner tries to acquire the time sequence characteristics of more passenger flow historical data by adopting a method of fusing multiple prediction models, so that higher prediction accuracy is achieved, but the improvement effect is not optimistic on the passenger flow data with strong fluctuation. In order to reduce the fluctuation of the passenger flow data, the original data can be filtered to be smooth, but the data is distorted, and finally the predicted value is greatly different from the actual value. The original passenger flow data is decomposed by data decomposition to obtain a plurality of subsequences, and the characteristics of nonlinearity and instability of the passenger flow data can be well processed by predicting the subsequences respectively, so that the method has important significance for realizing high-precision passenger flow prediction.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide a high-precision passenger flow prediction method, which can reduce prediction errors caused by strong and non-stationarity of passenger flow data and realize more accurate passenger flow prediction.
The technical scheme is as follows: a city rail station passenger flow prediction method based on CEEMDAN and BLSTM combined models comprises the following steps:
step 1: the method comprises the steps that a background management system of the urban rail station acquires historical passenger flow data of an in-out station within a period of time when the in-out station enters and exits the station;
step 2: preprocessing historical passenger flow data, intercepting time points in a station operation time period from the passenger flow data, and removing abnormal data;
and step 3: classifying the preprocessed historical passenger flow data according to the types of working days and holidays to respectively obtain passenger flow data of working day characteristics and holiday characteristics;
and 4, step 4: respectively analyzing the correlation between working day and holiday passenger flow historical data by adopting a Pearson correlation coefficient;
and 5: respectively carrying out data decomposition on the working day passenger flow data and the holiday passenger flow data by adopting CEEMDAN to obtain a plurality of IMF components and a residual component;
step 6: normalizing each component obtained in the step 5, establishing a sliding time window by using the first s adjacent historical data which are analyzed and obtained in the step 4 and show strong positive correlation, and then reconstructing the sliding time window into three-dimensional data;
and 7: establishing a BLSTM prediction model; the model consists of an input layer, a hidden layer and an output layer, wherein the input layer and the output layer adopt a full-connection layer, namely a Dense layer, the hidden layer is BLSTM, the number of neurons of the input layer and the hidden layer is determined through multiple experiments, and the number of neurons of the output layer is set to be 1;
and 8: and (4) sequentially sending the data obtained in the step (6) into a BLSTM model for prediction, performing inverse normalization on the prediction result of each component and storing the prediction result, and adding the prediction results after the prediction of all the components is finished to obtain a final passenger flow prediction value.
Further, in step 1, a period of time is divided at fifteen minute particle size.
Further, in step 2, if the passenger flow data in the initial time period and the last time period of the operation is less, discarding the part of data; if abnormal data exists in the passenger flow data of the station management system, the average value of the passenger flow volume of the abnormal data at the previous time and the next time or the data at the same time in the previous day are replaced.
Further, in step 4, the Pearson correlation coefficient formula is as follows:
Figure BDA0003472627090000021
in the formula, Yi，XiRespectively passenger flow and ith data of influencing factors,
Figure BDA0003472627090000031
respectively passenger flow data, average number of influencing factors, SY，SXRespectively passenger flow data and variance of influencing factors, r is a correlation coefficient, and the value range is [ -1, 1]The more r is greater than zero and closer to 1 the stronger the positive correlation, and vice versa.
Further, in step 6, the normalization processing formula is as follows:
Figure BDA0003472627090000032
in the formula, x' is normalized passenger flow data, x is original passenger flow data, and x ismax，xninThe maximum and minimum values of x, respectively; the normalized passenger flow data was set to 80% as training data and 20% as test data.
Further, the specific process of step 5 is:
step 5.1: adding noise to the passenger flow data;
x'(t)＝x(t)+ε0n(t) (3)
wherein x' (t) is the traffic data after white noise is added, x (t) is the original traffic data, ε0N (t) is white gaussian noise that follows an N (0, 1) distribution;
step 5.2: the decomposition of x' (t) using EMD yields a first order component imf11(t)；
Figure BDA0003472627090000033
Wherein T is the number of times of decomposition by EMD;
step 5.3: calculating the residual amount obtained by the first step of decomposition;
r1(t)＝x(t)-imf1(t) (5)
step 5.4: calculating the second order component imf2(t)，Ej[]Is the j-th order component of EMD decomposition of the data;
Figure BDA0003472627090000034
step 5.5: repeating the steps 5.3 and 5.4, calculating the residual quantity of the kth sequence component and calculating the kth +1 component according to the residual quantity;
rk(t)＝rk-1(t)-imfk(t) (7)
Figure BDA0003472627090000035
step 5.6: repeating the step 5.5 until the obtained residual signal quantity extreme points do not exceed two and K sequence components are obtained, wherein the final residual quantity is as follows:
Figure BDA0003472627090000041
step 5.7: the CEEMDAN decomposition results were obtained as:
Figure BDA0003472627090000042
further, the specific process of step 6 is:
taking the time point needing to predict the passenger flow as T0The time before the time is T1And so on; when constructing the sliding window, taking the time window s obtained in the step 4 as a time step,
by [ T ]n，Tn-1，Tn-2，...，Tn-s+1]Historical component data as input data, with Tn-sThe component data of the time as output data, with [ T ]n-1，Tn-2，Tn-3，...，Tn-s]Historical component data as input data, with Tn-s-1The component data of the moment is used as output data; constructing training data and test data into corresponding input and output data through sliding of time window, and using Ts，Ts-1，Ts-2，...，T1]Historical data prediction yields T0Time of day component data, will T0Adding the data of each component at the moment to obtain the final T0Temporal traffic prediction data.
Further, the step 7 prediction process has the calculation formula as follows:
forget the door:
ft＝σ(Wf·[st-1，xt]+bf) (11)
an input gate:
it＝σ(Wi·[st-1，xt]+bi) (12)
gt＝tanh(Wg[st-1，xt]+bg) (13)
Figure BDA0003472627090000043
an output gate:
ot＝σ(Wo[st-1，xt]+bo) (15)
Figure BDA0003472627090000044
forward propagation layer:
st＝f(w1·xt+w2·st-1) (17)
a counter-propagating layer:
s′t＝f(w3·xt+w5·s′t-1) (18)
and (3) final output:
ot＝g(w4·st+w6·s′t) (19)
in the formula, xtFor inputting data, stFor hidden layer output, ctFor the cell state, W, b are weighted and biased, σ and tanh represent sigmoid and tanh activation functions, respectively, and the equations for these are:
Figure BDA0003472627090000051
Figure BDA0003472627090000052
furthermore, the passenger flow data obtained by prediction by the method can be used as the basis for predicting the air conditioner load in the station together with the temperature and humidity related characteristics, and the energy-saving control of the air conditioner is further performed.
In particular implementations, there is a computer readable storage medium comprising one or more programs for execution by one or more processors, the one or more programs including instructions for performing the method of any of claims 1-9.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages:
(1) the passenger flow change characteristics of different working days and holidays are considered, historical passenger flow data of the working days and the holidays are separated and respectively predicted, interference caused by different passenger flow characteristics is avoided, and a reasonable sliding time window is determined by adopting a Pearson correlation coefficient, so that a prediction model can obtain more historical data characteristics;
(2) the CEEMDAN is adopted to decompose the original passenger flow data, and a plurality of subsequences obtained by decomposition are respectively predicted, so that the problem of difficult prediction caused by nonlinearity, non-stationarity and strong fluctuation of the original data is solved;
(3) and a BLSTM prediction model improved based on LSTM is adopted for prediction, and the model can extract the time sequence characteristics of the data in the positive and negative directions and has better prediction accuracy compared with other models.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a view of the internal structure of the LSTM;
FIG. 3 is a block diagram of a BLSTM structure;
FIG. 4 is a decomposition result of CEEMDAN workday passenger flow data;
FIG. 5 is a decomposition result of CEEMDAN holiday passenger flow data;
FIG. 6 shows the result of the forecast of the traffic on the working day of the method;
FIG. 7 shows the results of the method for predicting holiday passenger flow.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1 to 3, the urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model adopts a data decomposition algorithm and a prediction model to be combined to improve the passenger flow prediction accuracy, and specifically includes the following steps:
step 1: and acquiring historical passenger flow data of the station entering and exiting on a fifteen-minute time scale by the urban rail station background management system.
Step 2: and preprocessing the historical passenger flow data. And intercepting the time point of the obtained passenger flow data in the station operation time period, and discarding the part of data if the passenger flow data in the initial operation time period and the last operation time period are few. The passenger flow data obtained by the station management system may have extremely individual error data, and the obvious abnormal data is replaced by the average value of the passenger flow volume at the front time and the back time or the data at the same time in the previous day.
And step 3: and classifying the preprocessed historical passenger flow data according to the types of working days and holidays to respectively obtain passenger flow data of working day characteristics and holiday characteristics.
And 4, step 4: and (3) respectively analyzing the correlation between the working day and the historical data of the holiday passenger flow by adopting a Pearson correlation coefficient shown in the formula (1).
Figure BDA0003472627090000061
In the formula Yi，XiRespectively passenger flow and ith data of influencing factors,
Figure BDA0003472627090000062
respectively passenger flow data, average number of influencing factors, SY，SXRespectively passenger flow data and variance of influencing factors, r is a correlation coefficient, and the value range is [ -1, 1]The more r is greater than zero and closer to 1 the stronger the positive correlation, and vice versa.
And 5: respectively carrying out data decomposition on the working day passenger flow data and the holiday passenger flow data by adopting CEEMDAN to obtain a plurality of IMF components and a residual component, and specifically comprising the following steps:
1): adding noise to the passenger flow data;
x'(t)＝x(t)+ε0n(t) (2)
where x' (t) is the traffic data after white noise addition, x (t) is the original traffic data, ε0For noise standard deviation, N (t) is white gaussian noise that follows an N (0, 1) distribution.
2) The decomposition of x' (t) using EMD yields a first order component imf1(t)；
Figure BDA0003472627090000063
Wherein T is the number of times of decomposition by EMD;
3) calculating the residual amount obtained by the first step of decomposition;
r1(t)＝x(t)-imf1(t) (4)
4) calculating the second order component imf2(t)，Ej[]Is the j-th order component of EMD decomposition of the data;
Figure BDA0003472627090000071
5) repeating the third step and the fourth step, calculating the residual quantity of the kth sequence component and calculating the kth +1 component according to the residual quantity;
rk(t)＝rk-1(t)-imfk(t) (6)
Figure BDA0003472627090000072
6) repeating the fifth step until the obtained residual signal quantity extreme points do not exceed two and K sequence components are obtained, wherein the final residual quantity is as follows:
Figure BDA0003472627090000073
7) the CEEMDAN decomposition results were obtained as:
Figure BDA0003472627090000074
step 6: and (3) carrying out normalization processing on each component obtained in the step (5) by adopting an expression (10), taking 80% of the passenger flow data after the normalization processing as training data and 20% as test data, then establishing a sliding time window by using the previous s adjacent historical data which are analyzed and obtained in the step (4) and have stronger positive correlation, and then reconstructing the sliding time window into three-dimensional data.
Figure BDA0003472627090000075
Wherein x' is normalized passenger flow data, x is original passenger flow data, and x ismax，xminThe maximum and minimum values of x, respectively.
The specific steps of constructing the sliding time window are as follows:
taking the time point needing to predict the passenger flow as T0The time before the time is T1And so on. Taking the time window s obtained in the step 4 as a time step when constructing the sliding window,
firstly, with [ Tn，Tn-1，Tn-2，...，Tn-s+1]Historical component data as input data, with Tn-sThe component data of the time is used as output data, and then [ T ] is usedn-1，Tn-2，Tn-3，...，Tn-s]Historical component data as input data, with Tn-s-1The component data of the time of day is taken as output data. Constructing training data and test data into corresponding input and output data through sliding of a time window, and finally using [ T ]s，Ts-1，Ts-2，...，T1]Historical data prediction yields T0Time of day component data, will T0Adding the data of each component at the moment to obtain the final T0Temporal traffic prediction data.
And 7: and establishing a BLSTM prediction model. The model consists of an input layer, a hidden layer and an output layer, wherein the input layer and the output layer adopt a full connecting layer, namely a Dense layer, the hidden layer is BLSTM, the number of neurons of the optimal input layer and the hidden layer is determined through multiple experiments, and the number of neurons of the output layer is set to be 1. The specific calculation formula of the prediction process is as follows:
1) forget the door:
ft＝σ(Wf·[st-1，xt]+bf) (11)
2) an input gate:
it＝σ(Wi·[st-1，xt]+bi) (12)
gt＝tah(Wg[st-1，xt]+bg) (13)
Figure BDA0003472627090000081
3) an output gate:
ot＝σ(Wo[st-1，xt]+bo) (15)
Figure BDA0003472627090000082
4) forward propagation layer:
st＝f(w1·xt+w2·st-1) (17)
5) a counter-propagating layer:
s′t＝f(w3·xt+w5·s′t-1) (18)
6) and (3) final output:
ot＝g(w4·st+w6·s′t) (19)
in the formula xtFor inputting data, stFor hidden layer output, ctFor the cell state, W, b are weighted and biased, σ and tanh represent sigmoid and tanh activation functions, respectively, and the equations for these are:
Figure BDA0003472627090000083
Figure BDA0003472627090000084
and 8: and (4) sequentially sending the data obtained in the step (6) into a BLSTM model for prediction, performing inverse normalization on the prediction result of each component and storing the prediction result, and adding the prediction results after the prediction of all the components is finished to obtain a final passenger flow prediction value.
Examples
In the embodiment, the method is implemented through specific data, and results show that compared with the traditional passenger flow prediction method, the method can achieve a passenger flow prediction effect with higher precision, and can provide a passenger flow prediction value of future time with higher precision for subway operation departments. The data for this example are as follows:
the subway passenger flow data selected in the embodiment is passenger flow historical data of 44 days in total from 6 months 8 days to 7 months 22 days at a certain station of Guangxi Nanning 2020, wherein 31 days of working days and 13 days of holidays, passenger flow data from 6 points earlier to 11 points later are selected every day, the passenger flow data are divided by granularity of 15 minutes, missing values and bad data in the passenger flow data are preprocessed, the processed data are analyzed by using Pearson correlation coefficients, and model input data with time step lengths of 4 and 8 are respectively created for the passenger flow prediction models of the working days and the holidays by using a sliding window.
The prediction model parameters of the present embodiment are set as:
the model hidden layer is BLSTM, adopts the sense layer as input layer and output layer, and this embodiment sets up input neuron to 128, and hidden layer neuron number is 64, and the output step is 1, and output layer neuron number is 1 promptly, adds Dropout to hidden layer simultaneously and equals 0.3, prevents that the model from overfitting, and the activation function is tanh. Adam is selected as the optimizer, learning rate change setting is adopted, the initial learning rate is set to be 0.01, the learning rate is changed to be 1/10 after each 100 times of training, the training times are set to be 300 times, and the number of samples batch _ size input each time is set to be 50.
As can be seen from fig. 4 and 5, the cemdan decomposes the holiday passenger flow data and the weekday passenger flow data into 10 components and 7 components, respectively, and the fluctuation degree of each component gradually decreases, the first component fluctuates most severely, and the last component is a residual component obtained after decomposition.
As can be seen from fig. 6 and 7, the urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model provided by the invention can accurately predict the passenger flow volume in the future time period under the condition that the passenger flow data has strong fluctuation, the fitting effect of the predicted value and the actual value is good, and the predicted average percentage errors of the passenger flow data on the working days and the holidays are respectively 9.79% and 8.97%. Compared with the method, the average percentage errors of passenger flow data prediction of the working days directly by adopting the BLSTM prediction model without decomposing passenger flow data and by adopting the EMD and BLSTM combined prediction model are respectively 21.89% and 11.72%, and the EMD has the modal aliasing problem, so that the method can be proved to realize passenger flow prediction of subway stations with higher precision.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The embodiments of the present invention are not described in detail, but are known in the art, and can be implemented by referring to the known techniques.

Claims (10)

1. A method for predicting urban rail station passenger flow based on CEEMDAN and BLSTM combined models is characterized by comprising the following steps:
step 1: the method comprises the steps that a background management system of the urban rail station acquires historical passenger flow data of an in-out station within a period of time when the in-out station enters and exits the station;
step 2: preprocessing historical passenger flow data, intercepting time points in a station operation time period from the passenger flow data, and removing abnormal data;
and step 3: classifying the preprocessed historical passenger flow data according to the types of working days and holidays to respectively obtain passenger flow data of working day characteristics and holiday characteristics;
and 4, step 4: respectively analyzing the correlation between working day and holiday passenger flow historical data by adopting a Pearson correlation coefficient;
and 5: respectively carrying out data decomposition on the working day passenger flow data and the holiday passenger flow data by adopting CEEMDAN to obtain a plurality of IMF components and a residual component;
step 6: normalizing each component obtained in the step 5, establishing a sliding time window by using the first s adjacent historical data which are analyzed and obtained in the step 4 and show strong positive correlation, and then reconstructing the sliding time window into three-dimensional data;
and 7: establishing a BLSTM prediction model; the model consists of an input layer, a hidden layer and an output layer, wherein the input layer and the output layer adopt a full-connection layer, namely a Dense layer, the hidden layer is BLSTM, the number of neurons of the input layer and the hidden layer is determined through multiple experiments, and the number of neurons of the output layer is set to be 1;
and 8: and (4) sequentially sending the data obtained in the step (6) into a BLSTM model for prediction, performing inverse normalization on the prediction result of each component and storing the prediction result, and adding the prediction results after the prediction of all the components is finished to obtain a final passenger flow prediction value.
2. The method for predicting the passenger flow at the urban rail station based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein in the step 1, a period of time is divided in fifteen-minute granularity.
3. The method for predicting the passenger flow of the urban rail station based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein in the step 2, if the passenger flow data in the initial time period and the final time period of the operation is less, the part of the data is discarded; if abnormal data exists in the passenger flow data of the station management system, the average value of the passenger flow volume of the abnormal data at the previous time and the next time or the data at the same time in the previous day are replaced.
4. The method for predicting the passenger flow at the urban rail station based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein in the step 4, the Pearson correlation coefficient formula is as follows:
Figure FDA0003472627080000011
in the formula, Yi，XiRespectively passenger flow and ith data of influencing factors,
Figure FDA0003472627080000012
respectively passenger flow data, average number of influencing factors, SY，SXRespectively passenger flow data and variance of influencing factors, r is a correlation coefficient, and the value range is [ -1, 1]The more r is greater than zero and closer to 1 the stronger the positive correlation, and vice versa.
5. The urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein in step 6, the normalization processing formula is as follows:
Figure FDA0003472627080000021
in the formula, x' is normalized passenger flow data, x is original passenger flow data, and x ismax，xminThe maximum and minimum values of x, respectively; the normalized passenger flow data was set to 80% as training data and 20% as test data.
6. The urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein the specific process of step 5 is as follows:
step 5.1: adding noise to the passenger flow data;
x'(t)＝x(t)+ε0n(t) (3)
wherein x' (t) is the traffic data after white noise is added, x (t) is the original traffic data, ε0N (t) is white gaussian noise that follows an N (0, 1) distribution;
step 5.2: the decomposition of x' (t) using EMD yields a first order component imf1(t)；
Figure FDA0003472627080000022
Wherein T is the number of times of decomposition by EMD;
step 5.3: calculating the residual amount obtained by the first step of decomposition;
r1(t)＝x(t)-imf1(t) (5)
step 5.4: calculating the second order component imf2(t)，Ej[]Is the j-th order component of EMD decomposition of the data;
Figure FDA0003472627080000023
step 5.5: repeating the steps 5.3 and 5.4, calculating the residual quantity of the kth sequence component and calculating the kth +1 component according to the residual quantity;
rk(t)＝rk-1(t)-imfk(t) (7)
Figure FDA0003472627080000024
step 5.6: repeating the step 5.5 until the obtained residual signal quantity extreme points do not exceed two and K sequence components are obtained, wherein the final residual quantity is as follows:
Figure FDA0003472627080000031
step 5.7: the CEEMDAN decomposition results were obtained as:
Figure FDA0003472627080000032
7. the urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein the specific process of step 6 is as follows:
taking the time point needing to predict the passenger flow as T0The time before the time is T1And so on; when constructing the sliding window, taking the time window s obtained in the step 4 as a time step,
by [ T ]n,Tn-1,Tn-2,...,Tn-s+1]Historical component data as input data, with Tn-sThe component data of the time as output data, with [ T ]n-1,Tn-2,Tn-3,...,Tn-s]Historical component data as input data, with Tn-s-1The component data of the moment is used as output data; constructing training data and test data into corresponding input and output data through sliding of time window, and using Ts,Ts-1,Ts-2,...,T1]Historical data prediction yields T0Time of day component data, will T0Adding the data of each component at the moment to obtain the final T0Temporal traffic prediction data.
8. The urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein the step 7 prediction process calculation formula is as follows:
forget the door:
ft＝σ(Wf·[st-1,xt]+bf) (11)
an input gate:
it＝σ(Wi·[st-1,xt]+bi) (12)
gt＝tanh(Wg[st-1,xt]+bg) (13)
Figure FDA0003472627080000033
an output gate:
ot＝σ(Wo[st-1,xt]+bo) (15)
Figure FDA0003472627080000034
forward propagation layer:
st＝f(w1·xt+w2·st-1) (17)
a counter-propagating layer:
s't＝f(w3·xt+w5·s't-1) (18)
and (3) final output:
ot＝g(w4·st+w6·s't) (19)
in the formula, xtFor inputting data, stFor hidden layer output, ctFor the cell state, W, b are weighted and biased, σ and tanh represent sigmoid and tanh activation functions, respectively, and the equations for these are:
Figure FDA0003472627080000041
Figure FDA0003472627080000042
9. the urban rail station passenger flow prediction method based on the CEEMDAN and BLSTM combined model as claimed in claim 1, wherein the passenger flow data obtained by the prediction method can also be used as the basis for predicting the air conditioner load in the station together with the temperature and humidity related characteristics to further perform energy-saving control on the air conditioner.
10. A computer readable storage medium comprising one or more programs for execution by one or more processors, the one or more programs comprising instructions for performing the method of any of claims 1-9.
CN202210047361.6A 2022-01-17 2022-01-17 Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model Pending CN114492967A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210047361.6A CN114492967A (en) 2022-01-17 2022-01-17 Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210047361.6A CN114492967A (en) 2022-01-17 2022-01-17 Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model

Publications (1)

Publication Number Publication Date
CN114492967A true CN114492967A (en) 2022-05-13

Family

ID=81511193

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210047361.6A Pending CN114492967A (en) 2022-01-17 2022-01-17 Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model

Country Status (1)

Country Link
CN (1) CN114492967A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115854501A (en) * 2023-01-10 2023-03-28 大连理工大学 Airport terminal room temperature large-lag prediction control method based on passenger flow prediction

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115854501A (en) * 2023-01-10 2023-03-28 大连理工大学 Airport terminal room temperature large-lag prediction control method based on passenger flow prediction
CN115854501B (en) * 2023-01-10 2024-03-19 大连理工大学 Airport terminal room temperature large hysteresis prediction control method based on passenger flow prediction

Similar Documents

Publication Publication Date Title
CN111301426B (en) Method for predicting energy consumption in future driving process based on GRU network model
Zhang et al. Traffic Accident Prediction Based on LSTM‐GBRT Model
CN111931978A (en) Urban rail transit passenger flow state prediction method based on space-time characteristics
CN111860989B (en) LSTM neural network short-time traffic flow prediction method based on ant colony optimization
CN107978153B (en) A kind of multimode traffic factors influencing demand analysis method based on space vector autoregression model
Lu et al. Short-term demand forecasting for online car-hailing using ConvLSTM networks
CN112765894B (en) K-LSTM-based aluminum electrolysis cell state prediction method
CN110459056A (en) A kind of public transport arrival time prediction technique based on LSTM neural network
CN115358491A (en) Subway passenger flow prediction method and system based on CEEMDAN-LSTM
CN116663742B (en) Regional capacity prediction method based on multi-factor and model fusion
CN114912666A (en) Short-time passenger flow volume prediction method based on CEEMDAN algorithm and attention mechanism
CN113205223A (en) Electric quantity prediction system and prediction method thereof
CN112668822A (en) Scientific and technological achievement transformation platform sharing system, method, storage medium and mobile phone APP
CN111277710A (en) Telephone traffic prediction method, system, computer device and storage medium
CN115496257A (en) Short-term vehicle speed prediction based on space-time fusion
CN115756922A (en) Fault prediction diagnosis method and device, electronic equipment and storage medium
CN115269758A (en) Passenger-guidance-oriented road network passenger flow state deduction method and system
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
CN116822652A (en) Subway fault prediction method, subway fault prediction device, electronic equipment, subway fault prediction system and storage medium
CN115730744A (en) Water consumption prediction method and system based on user mode and deep learning combined model
CN114331019A (en) Urban traffic safety risk real-time assessment method and device based on risk factor
CN114492967A (en) Urban rail station passenger flow prediction method and medium based on CEEMDAN and BLSTM combined model
CN117436653A (en) Prediction model construction method and prediction method for travel demands of network about vehicles
CN116128160B (en) Method, system, equipment and medium for predicting peak passenger flow of railway station
Sim et al. Temporal Attention Gate Network With Temporal Decomposition for Improved Prediction Accuracy of Univariate Time-Series Data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination